Advanced Microeconomics

Leonardo Felli
EC441: Room S.421
Lecture 2: 11 October 2006
Leonardo Felli Lecture 2: 11 October 2006
Utility Maximization
Up to now we focused on how to represent the consumers preferences.
We shall now consider the more sour note of the constraint that is imposed on
the consumer.
Denition of the budget set.
B(p, m) = {x | (p x) m, x X}
Slide 1
Leonardo Felli Lecture 2: 11 October 2006
-
6
x
1
x
2
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
B(p, m)
L = 2 X = R
2
+
Slide 2
Leonardo Felli Lecture 2: 11 October 2006
The two exogenous variables that characterize the consumers budget set are:
the level of income m
the vector of prices p = (p
1
, . . . , p
L
).
Often the budget set is characterized by a level of income represented by the
value of the consumers endowment x
0
(labour supply):
m = (p x
0
)
Slide 3
Leonardo Felli Lecture 2: 11 October 2006
The basic consumers problem (rational, continuous and monotonic preferences):
max
{x}
u(x)
s.t. x B(p, m)
Result 1: If p > 0 and u() is continuous, then the utility maximization
problem has a solution.
Proof: If p > 0 (meaning p
l
> 0, l = 1, . . . , L) the budget set is
compact (closed, bounded) hence by Weierstrass theorem the maximization of a
continuous function on a compact set has a solution.
Slide 4
Leonardo Felli Lecture 2: 11 October 2006
If u() is continuously dierentiable, the solution to the consumers problem
x

= x(p, m)
is characterized by the following necessary conditions:
there exists a Lagrange multiplier such that:
u(x

) p
and
x

[u(x

) p] = 0
where
u(x

) = [u
1
(x

), . . . , u
L
(x

)].
Slide 5
Leonardo Felli Lecture 2: 11 October 2006
Moreover
p x

m
and
[p x

m] = 0.
Meaning that l = 1, . . . , L:
u
l
(x

) p
l
and
x

l
[u
l
(x

) p
l
] = 0
Slide 6
Leonardo Felli Lecture 2: 11 October 2006
In other words, if x

l
> 0 then
u
l
(x

) = p
l
Moreover
L

l=1
p
l
x

l
m, and
_
L

l=1
p
l
x

l
m
_
= 0
In other words if > 0 (always true with monotonic preferences) then
L

l=1
p
l
x

l
= m.
Slide 7
Leonardo Felli Lecture 2: 11 October 2006
In the case L = 2 and X = R
2
+
these conditions are (graphic representation):
if x

1
> 0 and x

2
> 0 then
u
1
u
2
=
p
1
p
2
if
u
1
u
2
<
p
1
p
2
then x

1
= 0 and x

2
> 0
if
u
1
u
2
>
p
1
p
2
then x

1
> 0 and x

2
= 0
Slide 8
Leonardo Felli Lecture 2: 11 October 2006
The conditions we stated are merely necessary.
What about sucient conditions?
If u() is quasi-concave and monotone
u(x) = 0 for all x X,
then the Kuhn-Tucker rst order conditions are sucient.
Slide 9
Leonardo Felli Lecture 2: 11 October 2006
If u() is not quasi-concave then
a u() locally quasi-concave at x

(x

satisfying FOC)
will suce for a local maximum.
Local (strict) quasi-concavity can be veried by checking whether the determi-
nants of the bordered leading principal minors of order
r = 1, . . . , L
of the Hessian matrix of u() at x

have the sign of

2
u
x
2
1


2
u
x
1
x
L
.
.
.
.
.
.
.
.
.

2
u
x
1
x
L


2
u
x
2
L
_
_
_
_
_
The bordered Hessian of the utility function u() is:
_
H, [u(x

)]
T
u(x

) 0
_
Slide 11
Leonardo Felli Lecture 2: 11 October 2006
The bordered leading principal minor of order r of the Hessian is:
_
H
r
, [u(x

)]
T
r
[u(x

)]
r
0
_
where H
r
is the leading principal minor of order r.
In the case L = 2 the second order conditions hold whenever
d
dx
1
_
dx
2
dx
1
_
> 0 provided that
u
1
u
2
=
p
1
p
2
.
Slide 12
Leonardo Felli Lecture 2: 11 October 2006
The solution to the utility maximization problem are:
x = x(p, m) =
_
_
_
_
x
1
(p
1
, . . . , p
L
, m)
.
.
.
x
L
(p
1
, . . . , p
L
, m)
_
_
_
_
These are known as the Marshallian or uncompensated demand functions.
Notice that strong monotonicity of preferences implies that the budget constraint
will be binding when computed at the value of the Marshallian demands. (building
block of Walras Law)
Slide 13
Leonardo Felli Lecture 2: 11 October 2006
Denition: the function obtained by substituting the Marshallian demands in
the consumers utility function is the indirect utility function:
V (p, m) = u(x

(p, m))
Properties of the indirect utility function:
1.
V
m
0 and
V
p
i
0 for every i = 1, . . . , L.
2. V (p, m) continuous in (p, m).
Not proved, however it rules out situations in which the consumption feasible set
is non-convex (e.g. indivisibility).
Slide 14
Leonardo Felli Lecture 2: 11 October 2006
3. V (p, m) homogeneous of degree 0 in (p, m).
Denition: F(x) is homogeneous of degree r if and only if
F(k x) = k
r
F(x) k R
+
Proof: notice that if we multiply both the vector of prices p and the level of
income m by the same positive scalar R
+
we obtain the budget set:
B( p, m) = {x X | p x m} = B(p, m)
therefore the budget set is unaltered, hence the indirect utility function (and the
Marshallian demand) are unaltered.
Slide 15
Leonardo Felli Lecture 2: 11 October 2006
4. V (p, m) is quasi-convex in p, that is:
{p | V (p, m) k}
is a convex set for every k R.
Proof: let p, m and p

, be such that:
V (p, m) k V (p

, m) k.
Denote p

= tp + (1 t)p

for some 0 < t < 1.

, m) k.
Slide 16
Leonardo Felli Lecture 2: 11 October 2006
Dene:
B = {x | (p x) m}
B

= {x | (p

x) m}
B

= {x | (p

x) m}
Claim:
B

B B

Consider x B

, then
p

x = [tp + (1 t)p

] x
= t (p x) + (1 t) (p

x) m
which implies either p x m or/and p

x m, or x B B

.
Slide 17
Leonardo Felli Lecture 2: 11 October 2006
Now
V (p

, m) = max
{x}
u(x) s.t. x B

max
{x}
u(x) s.t. x B B

= max {V (p, m), V (p

, m)} k
Since by assumption: V (p, m) k and V (p

, m

) k.
Slide 18
Leonardo Felli Lecture 2: 11 October 2006
Properties of the Marshallian demand x(p, m):
1. x(p, m) is continuous in (p, m), (consequence of the convexity of prefer-
ences).
2. x
i
(p, m) homogeneous of degree 0 in (p, m).
Proof: once again if we multiply (p, m) by > 0:
B( p, m) = {x X | p x m} = B(p, m)
therefore the solution to the utility maximization problem is unaltered.
Slide 19
Leonardo Felli Lecture 2: 11 October 2006
Constrained Envelope Theorem
Consider the problem:
max
x
f(x)
s.t. g(x, a) = 0
The Lagrangian is:
L(x, , a) = f(x) g(x, a)
Slide 20
Leonardo Felli Lecture 2: 11 October 2006
The necessary FOC are:
f

(x

g(x

, a)
x
= 0
and
g(x

(a), a) = 0
Substituting x

(a) and

(a) in the Lagrangian we get:

(a))

(a) g(x

(a), a)
Slide 21
Leonardo Felli Lecture 2: 11 October 2006
Dierentiating we get:
dL(a)
d a
=
_
f

(x

g(x

, a)
x
_
d x

(a)
d a

g(x

(a), a)
d

(a)
da

(a)
g(x

, a)
a
=

(a)
g(x

, a)
a
by the necessary FOC.
In other words: to the rst order only the direct eect of a on the Lagrangian
function matters.
Slide 22
Leonardo Felli Lecture 2: 11 October 2006
3. Roys identity:
x
i
(p, m) =
V/p
i
V/m
By the constrained envelope theorem and the observation that:
V (p, m) = u(x(p, m)) (p, m) [p x(p, m) m]
we shall obtain:
V/p
i
= (p, m) x
i
(p, m) 0
Slide 23
Leonardo Felli Lecture 2: 11 October 2006
and
V/m = (p, m) 0
which is the marginal utility of income.
Notice: the sign of the two inequalities above prove property 1 of the indirect
utility function V (p, m).
We conclude the proof substituting
V/m = (p, m)
into
V/p
i
= (p, m) x
i
(p, m)
and solving for x
i
(p, m).
Slide 24
Leonardo Felli Lecture 2: 11 October 2006
4. Adding up results.
From the identity:
p x(p, m) = m p, m
Dierentiation with respect to p
j
gives:
x
j
(p, m) +
L

i=1
p
i
x
i
p
j
= 0
Slide 25
Leonardo Felli Lecture 2: 11 October 2006
or, more interestingly, with respect to m gives:
L

i=1
p
i
x
i
m
= 1
There does not exist a clear cut comparative-static property except:
0
L

i=1
p
i
x
i
p
h
= x
h
(p, m)
which means that at least one of the Marshallian demand function has to be
downward sloping in p
h
.
Slide 26
Leonardo Felli Lecture 2: 11 October 2006
Eect of a change in income on the level of the Marshallian demand:
x
l
m
In the two commodities graph the set of tangency points for dierent values of
m is known as the income expansion path.
In the commodity/income graph the set of optimal choices of the quantity of the
commodity is known as Engel curve.
Slide 27
Leonardo Felli Lecture 2: 11 October 2006
We shall classify commodities with respect to the eect of changes in income in:
normal goods:
x
l
m
> 0
neutral goods:
x
l
m
= 0
inferior goods:
x
l
m
< 0
Slide 28
Leonardo Felli Lecture 2: 11 October 2006
Notice that from the adding up results above for every level of income m at least
one of the L commodities is normal:
L

l=1
p
l
x
l
m
= 1
We shall classify commodities depending on the curvature of the Engel curves:
if the Engel curve is convex we are facing a luxury good
If the Engel curve is concave we are facing a necessity.
Slide 29
Leonardo Felli Lecture 2: 11 October 2006
-
6
x(p, m)
m
luxury
necessity
Slide 30
Leonardo Felli Lecture 2: 11 October 2006
Expenditure Minimization Problem
The dual problem of the consumers utility maximization problem is the expen-
diture minimization problem:
min
{x}
p x
s.t. u(x) U
Dene the solution as:
x = h(p, U) =
_
_
_
_
h
1
(p
1
, . . . , p
L
, U)
.
.
.
h
L
(p
1
, . . . , p
L
, U)
_
_
_
_
the Hicksian (compensated) demand functions.
Slide 31
Leonardo Felli Lecture 2: 11 October 2006
We shall also dene:
e(p, U) = p h(p, U)
as the expenditure function.
Properties of the expenditure function:
1. Continuous in p and U.
2.
e
U
> 0 (2.1) and
e
p
l
0 (2.2) for every l = 1, . . . , L.
Proof: (2.1): Suppose not.
Slide 32
Leonardo Felli Lecture 2: 11 October 2006
Then there exist U

< U

such that (denote x

and x

the corresponding
solution to the e.m.p.)
p x

p x

> 0
If the latter inequality is strict we have an immediate contradiction of x

solving
e.m.p.;
if on the other hand
p x

= p x

> 0
then by continuity and strict monotonicity of u() there exists (0, 1) close
enough to 1 such that
u( x

) > U

Slide 33
Leonardo Felli Lecture 2: 11 October 2006
and
p x

> p x

which contradicts x

solving e.m.p.
(2.2): consider p

and p

such that p

l
p

l
but p

k
= p

k
for every k = l.
Let x

and x

be the solutions to the e.m.p. with p

and p

respectively.
Then by denition of e(p, U)
e(p

, U) = p

= e(p

, U).
Slide 34
Leonardo Felli Lecture 2: 11 October 2006
3. Homogeneous of degree 1 in p.
Proof: The feasible set of the e.m.p. does not change when prices are multiplied
by the factor k > 0:
u(x) U
Hence k > 0, minimizing (k p) x on this set leads to the same answer.
Let x

be the solution, then:

= k e(p, U).
Slide 35
Leonardo Felli Lecture 2: 11 October 2006
4. Concave in p (graphic intuition).
Proof: let p

= t p + (1 t) p

for t [0, 1].

be the solution to e.m.p. for p

.
Then
e(p

, U) = p

= t p x

+ (1 t) p

t e(p, U) + (1 t) e(p

, U)
by denition of e(p, U) and e(p

, U) and u(x

) U.
Slide 36
Leonardo Felli Lecture 2: 11 October 2006
Properties of the Hicksian demand functions:
h(p, U)
1. Shephards Lemma.
e(p, U)
p
l
= h
l
(p, U)
Proof: by constrained envelope theorem.
Slide 37
Leonardo Felli Lecture 2: 11 October 2006
2. Homogeneity of degree 0 in p.
Proof: by Shephards lemma and the following theorem.
Theorem. If a function F(x) is homogeneous of degree r in x then (F/x
l
)
is homogeneous of degree (r 1) in x for every l = 1, . . . , L.
Proof: Dierentiating with respect to x
l
the identity that denes homgeneity
of degree r:
F(k x) = k
r
F(x) k > 0
Slide 38
Leonardo Felli Lecture 2: 11 October 2006
we obtain:
k
F(k x)
x
l
= k
r
F(x)
x
l
This is the denition of homogeneity of degree (r 1):
F(k x)
x
l
= k
(r1)
F(x)
x
l
.
Euler Theorem: If a function F(x) is homogeneous of degree r in x then:
r F(x) = F(x) x
Slide 39
Leonardo Felli Lecture 2: 11 October 2006
Proof: Dierentiating with respect to k the identity:
F(k x) = k
r
F(x) k > 0
we obtain:
F(kx) x = rk
(r1)
F(x)
for k = 1 we obtain:
F(x) x = r F(x).
Slide 40
Leonardo Felli Lecture 2: 11 October 2006
3. The matrix of own and cross-partial derivatives with respect to p (Substitution
matrix)
S =
_
_
_
_
h
1
p
1

h
1
p
L
.
.
.
.
.
.
.
.
.
h
L
p
1

h
L
p
L
_
_
_
_
is negative semi-denite and symmetric. (Main diagonal non-positive).
Proof: Simmetry follows from Shephards lemma and Young Theorem.
Indeed:
h
l
p
i
=

p
i
_
e(p, U)
p
l
_
=

p
l
_
e(p, U)
p
i
_
=
h
i
p
l
Slide 41
Leonardo Felli Lecture 2: 11 October 2006
While negative semi-deniteness follows from the concavity of e(p, U) and the
observation that S is the Hessian of the function e(p, U).
Identities:
V [p, e(p, U)] U
x
l
[p, e(p, U)] h
l
(p, U) l
e[p, V (p, m)] m
h
l
[p, V (p, m)] x
l
(p, m) l
Slide 42
Leonardo Felli Lecture 2: 11 October 2006
Slutsky decomposition:
Start from the identity
h
l
(p, U) x
l
[p, e(p, U)]
if the price p
i
changes the eect is:
h
l
p
i
=
x
l
p
i
+
x
l
m
e
p
i
Slide 43
Leonardo Felli Lecture 2: 11 October 2006
Notice that by Shephards lemma:
e
p
i
= h
i
(p, U) = x
i
[p, e(p, U)]
then
h
l
p
i
=
x
l
p
i
+
x
l
m
x
i
.
or
x
l
p
i
=
h
l
p
i

x
l
m
x
i
.
Slide 44
Leonardo Felli Lecture 2: 11 October 2006
Own price eect gives Slutsky equation:
x
l
p
l
=
h
l
p
l

x
l
m
x
l
.
Substitution eect:
h
l
p
l
Income eect:
x
l
m
x
l
Slide 45
Leonardo Felli Lecture 2: 11 October 2006
x
2
6
x
1
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e
e -
s
s
y

x
0
x
1
x
2
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
s
Slide 46
Leonardo Felli Lecture 2: 11 October 2006
We know the sign of the substitution eect it is non-positive.
The sign of the income eect depends on whether the good is normal or inferior.
In the case that:
x
l
p
l
> 0
we conclude that the good is Gien.
This is not a realistic feature: inferior good with a big income eect.
Slide 47